#asahi-dev · 2022-02-17 04:53
marcanif it violates barrier guarantees in a way that can cause proper corruption when you yank the plug on the mac mini, we repro it then file a bug then I blog about it so it hits hacker news ;)#asahi-dev · 2022-02-17 04:53
marcanthe question is what does macOS do#asahi-dev · 2022-02-17 04:53
marcan< sven> we can always file a radar with apple and have it disappear and get no feedback for years! ;)#asahi-dev · 2022-02-16 18:35
marcanthat needs to go in the DT#asahi-dev · 2022-02-16 18:35
marcanat the very least the DCP DVA offset is different#asahi-dev · 2022-02-16 18:35
marcanI imagine it won't work if nobody has tried it#asahi-dev · 2022-02-16 18:31
marcanI want to merge that if it's usable on all the machines these days too, so that will be useful#asahi-dev · 2022-02-16 18:29
marcanI kind of hate it but...#asahi-dev · 2022-02-16 18:29
marcandid you look at the atomic stuff?#asahi-dev · 2022-02-16 18:29
marcanyup#asahi-dev · 2022-02-16 18:28
marcanI'll clean up kernel branches tomorrow, today I ended up spending most of the day on non-asahi stuff#asahi-dev · 2022-02-16 18:28
marcananyway, I should get some sleep#asahi-dev · 2022-02-16 18:27
marcanalso note that this is in japan, which has the lowest AC voltage in the world, so worst possible case :)#asahi-dev · 2022-02-16 18:26
marcanturns out when your entire system sips power, that PSU main reservoir cap is pretty good...#asahi-dev · 2022-02-16 18:25
marcanwho needs SSD backup caps when the entire system is low power and the PSU is built in? that's a free 2 second UPS.#asahi-dev · 2022-02-16 18:24
marcanso if there is no actual last gasp signal, that is one hack of a missed opportunity :p#asahi-dev · 2022-02-16 18:24
marcanthat said, that thing keeps running for like >2 seconds after I pull the plug#asahi-dev · 2022-02-16 18:23
marcanI need to look closer but I didn't see anything obvious going to nvme right now#asahi-dev · 2022-02-16 18:23
marcanOTOH on the mac mini, if there is no last gasp signal, this means potential data loss#asahi-dev · 2022-02-16 18:23
marcanat least for laptops this is a ~non issue, inherently, especially if we can make sure a flush gets issued on panic and other special cases like that#asahi-dev · 2022-02-16 18:22
marcan(in intervals9#asahi-dev · 2022-02-16 18:22
marcansven: maybe just looking at flush commands from the HV and seeing if there is some logical distribution?#asahi-dev · 2022-02-16 18:21
marcanI need to get some sleep, but yeah#asahi-dev · 2022-02-16 18:21
marcanI wouldn't put it past apple to do some similar hack#asahi-dev · 2022-02-16 18:20
marcanI wonder what macos does#asahi-dev · 2022-02-16 18:20
marcaneven 1 second tbh#asahi-dev · 2022-02-16 18:20
marcan:)#asahi-dev · 2022-02-16 18:20
marcanalso yeah, now I remember it was that *plus* I had an LD_PRELOAD to stop firefox from issuing syncs all the time and breaking it#asahi-dev · 2022-02-16 18:20
marcanha#asahi-dev · 2022-02-16 18:19
marcanhaven't seen any last-gasp stuff from macos#asahi-dev · 2022-02-16 18:19
marcan< axboe> "only issue flush if X seconds has passed since last issue" <- isn't that what laptop_mode used to do back in the day?#asahi-dev · 2022-02-16 18:07
marcanyeah but that normally wouldn't happen#asahi-dev · 2022-02-16 18:06
marcanlet's see#asahi-dev · 2022-02-16 18:06
marcanI should test if macos does anything funky on the mac mini when pulling the power#asahi-dev · 2022-02-16 18:05
marcanaxboe: I have a suspicion that macos doesn't issue flushes and instead relies on having the proper last-gasp hooks to guarantee it'll issue one (and it will complete) if power is lost, or maybe that's outright guaranteed by the controller itself?#asahi-dev · 2022-02-15 23:01
marcanthe fix will only work if only core #3 is missing from each cluster#asahi-dev · 2022-02-15 23:00
marcanwe actually need to renumber the nodes in cpu-map, since the code expects them to be consecutive#asahi-dev · 2022-02-15 23:00
marcanjannau, axboe: I expected the cpu-map issue; also that fix is probably incorrect#asahi-dev · 2022-02-15 15:42
marcanalso the RTC stuff is, of course, >12.x only#asahi-dev · 2022-02-15 15:41
marcanjust needs the nvmem offsets, which I think I pasted here previously#asahi-dev · 2022-02-15 15:41
marcanyeah, mostly#asahi-dev · 2022-02-15 15:39
marcanthanks :)#asahi-dev · 2022-02-15 15:39
marcant6000 should work; t8103 needs devicetree stuff (the nvmem/spmi bits)#asahi-dev · 2022-02-15 15:38
marcanok, too tired to do more work today, I'm going to sleep early for once. pushed things as is to spmi/work.#asahi-dev · 2022-02-13 15:22
marcankettenis: I think that's the power on after AC loss flag, but I haven't tested it yet#asahi-dev · 2022-02-13 15:15
marcanso I think we're pretty safe#asahi-dev · 2022-02-13 15:15
marcanand writing to it won't brick anything; I tried :p#asahi-dev · 2022-02-13 15:15
marcanthe RTC stuff shouldn't have any side effects when read#asahi-dev · 2022-02-13 15:14
marcanbut FWIW I've been messing with nvram a lot today and nothing weird happened#asahi-dev · 2022-02-13 15:14
marcan*writing* to files in sysfs is a bit different from rm :p